Fraud & Abuse Prevention (a.k.a Community Trust) : • Protect shoppers from inappropriate and untrustworthy community contributors and contents (e.g., Reviews, Q&A, Votes)• Build systems to detect abuse at scale and share the decisions with other teams• Tech Stack: AWS, Java, Guice, Akka, SQLContributions:• Lead retrospective meetings of the team, drive discussions to improve the development processes

Community TrustContributions:• Designed and implemented a data backfilling workflow to migrate ~31 billion records (~12 TB) from the old data store to the new one, tackled complex problems including: (1) cleaning malformed legacy data, (2) exploring 10+ data migration tools, (3) reducing the estimated cost for backfilling by ~75%, (4) coordinating the migration project with other ongoing projects• Completed the data migration seamlessly with no system downtime, i.e., the system continuously served clients and processed live traffic during the migration• Proposed and implemented several data deletion solutions so that our data stores conform with the General Data Protection Regulation• Hosted Q&A sessions (both in-person and virtual) for interns• Lead retrospective meetings of the team, drive discussions to improve the development processes

• Designed and implemented a benchmarking framework to automate code testing and deployment• Researched computer vision related algorithms, implemented the benchmarking logic in Python• Wrote YAML files to orchestrate and manage cloud resources• Developed Docker files to containerize the benchmarking source code package

• Identified 12 bugs in central management system (CMS) of security cameras with integration test in C++• Tested basic operations (pan, tilt and zoom) and recording functions of 7 cameras• Corrected roughly 10 mistakes in CMS User Manual, both technical and grammatical

• Designed and 3D-print an electronic part sample• Established partnership with 2 companies• Completed 4 transactions with 4 different companies with gross profit of over $60,000

Research: 3D Scanning and Reconstruction with Intel RealSense Camera• Researched, implemented, and improved 25 algorithms relevant to image filtering, feature extraction, point cloud registration, and surface reconstruction (largely in C++)• Developed open-source software for 3D reconstruction without proprietary APIs• Awarded “Remarkable Research Prize of Tsinghua University” (top 2%)

Research: Web-video Classification System Based on Audio and Visual Information• Wrote scripts to preprocess web videos and extract over 20 kinds of audio and visual features, including Mel-frequency cepstral coefficients (MFCCs) and color histograms• Trained a convolutional neural network (CNN) to classify videos• Ranked first out of nearly 70 systems developed by other students in F-score on test sets